Real-Time Measurement of the Vertical Binding Energy during the

Jan 22, 2015 - Using femtosecond time-resolved two-photon photoelectron spectroscopy, we determine (i) the vertical binding energy (VBE = 0.8 eV) of ...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/JACS

Real-Time Measurement of the Vertical Binding Energy during the Birth of a Solvated Electron Julia Staḧ ler,* Jan-Christoph Deinert, Daniel Wegkamp, Sebastian Hagen, and Martin Wolf Abteilung Physikalische Chemie, Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14195 Berlin, Germany ABSTRACT: Using femtosecond time-resolved two-photon photoelectron spectroscopy, we determine (i) the vertical binding energy (VBE = 0.8 eV) of electrons in the conduction band in supported amorphous solid water (ASW) layers, (ii) the time scale of ultrafast trapping at pre-existing sites (22 fs), and (iii) the initial VBE (1.4 eV) of solvated electrons before significant molecular reorganization sets in. Our results suggest that the excess electron dynamics prior to solvation are representative for bulk ASW.



terahertz (THz) spectroscopy,14 photoexcitation to the CB first creates a delocalized excess electron, (H2O)−deloc (dark blue curve). This excess electron15 gains considerable binding energy by localization in the solvation shell, (H2O)loc, and thereby compensates λaq; the total energy is minimized at q0. The vertical binding energy of such relaxed SEs, VBEaq, was determined by TR photoelectron (PE) spectroscopy for both liquid water (bulk, 3.3 eV; surface, 1.6 eV)16,17 and ice crystallites (surface, 3.8 eV).18,19 Despite numerous studies, however, the VBE of the water CB remains unknown. The reason for this is the ultrashort lifetime of electrons in this delocalized state, as localization through electron solvation is energetically much more favorable. Due to the rapidness of this process, neither the localization dynamics nor the mechanism (small polaron formation by self-trapping versus localization at pre-existing potential minima) could be resolved until now. Besides time resolution, one main experimental challenge for this particular question is the controlled injection of electrons into the water CB and their subsequent probing by TR spectroscopy, which can be solved by photoexcitation of, e.g., H2O itself or ions in solution.13,17 However, the formation dynamics of SEs are influenced by the presence of the positively charged donor. A complementary approach that does not involve positively charged ions in the vicinity of the SE involves the injection of electrons from a metallic or semiconducting substrate. This can be performed at solid−liquid interfaces4,5 and also in adsorbed amorphous solid water (ASW) layers.20−23 It was shown that the excess electrons are transferred to a localized SE state, eS. Rearrangement of the water molecules leads to a continuous binding energy increase, while the electron−ice complex moves toward its new equilibrium position and its population reduces due to competing decay to the metal substrate.21,23 These SEs are localized in the second−third ice bilayer (BL) in front of the metal template.22 Despite these previous studies, (i) the VBE of the ice CB, (ii)

INTRODUCTION Excess electrons in aqueous environments play a crucial role in physics, chemistry, and biology, not least due to the importance of water as the paramount solvent in nature.1,2 The alignment of energy levels of solute molecules with respect to the affinity level of water plays a crucial role in charge-transfer reactions, as it determines the probability of solvated electron (SE) formation. Thereby, an excess electron is stabilized by the polar, aqueous surroundings. SEs are known to be highly reactive species, for example, in dissociative electron attachment,3 ammonia synthesis,4 and CO2 reduction in water5 or ice6 solutions. One way of looking at the energy levels of water is by interpreting it as an amorphous large-band-gap semiconductor with an occupied valence band (VB) and unoccupied conduction band (CB). The transport of electrons is then determined by the band dispersion and competing decay channels, like localization and trapping of the charge carriers. A negatively charged electron−water complex can then be considered to be an anionic defect of pure water.7 The injection of electrons requires energetic resonance of the donor level with the water CB; for electron-induced chemical reactions, the relaxed water−electron complex (H2O)− must be in resonance with accepting states of the reactant. The vertical binding energies (VBEs) of these energy levels are difficult to determine directly for liquid water. This led to numerous experimental and theoretical studies on water anion clusters in the gas phase,8−12 aiming at the extrapolation of VBEs to the bulk value. When discussing the solvation dynamics of excess electrons in an aqueous environment, it is helpful to consider the energy levels in a Marcus type of picture (cf. Figure 1a), where the total energy consists of the binding energy of the electron and the reorganization energy stored in the distorted solvent. In the case of neutral water (H2O, orange curve), the electron is still in the VB, and formation of the (empty) solvation shell requires the reorganization energy λaq. As shown by scavenger experiments13 and, very recently, by time-resolved (TR) © XXXX American Chemical Society

Received: November 17, 2014

A

DOI: 10.1021/ja511571y J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 1. (a) Marcus parabolas for solvated electrons in water/ice and time-resolved photoelectron spectroscopy scheme. (b) Angle-integrated twophoton photoelectron intensity of amorphous solid water on Cu(111) as a function of energy and pump−probe time delay. (c) Indicated elementary processes: (1) metal electron injection into the ice conduction band, (2) relaxation to the band minimum on fs time scales, (3) eS population, and (4) energetic stabilization and population decay to the metal substrate. function of the sample are known, the PE intensity distribution can be plotted as a function of initial state energy below the Fermi energy, Einitial − EF = Ekin + Φ − (hν1 + hν2), intermediate state energy above it, Einter − EF = Ekin + Φ − hν2, or VBE = hν2 − Ekin. The energy resolution of the experiment is 60 meV.

the time scale of electron localization, and (iii) the fundamental question of whether the excess electrons get trapped by small polaron formation or if they localize at pre-existing trapping sites remain unanswered. In this Article, we present femtosecond TR two-photon photoelectron (2PPE) experiments with significantly improved time resolution compared to previous work that allows for a clear assignment of all involved processes on the electron’s passage through the ice CB of ASW grown on a Cu(111) template. Figure 1c illustrates all the elementary processes: The electrons are photoexcited from a (through D2O adsorption) strongly broadened Cu(111) surface state (SS) into the continuum of ice CB states (1), before they rapidly relax with a remarkably high rate of 4 meV/fs toward the bottom of the CB (2), which lies at a VBE of 0.8 eV. The carriers are then trapped in pre-existing, localized traps with a characteristic time constant of 22 fs (3) and are subsequently stabilized energetically by rearrangement of the ice network (4). This first real-time measurement of excess electron dynamics in and out of the CB of an aqueous solvent provides unprecedented knowledge of the initial steps of SE formation.





RESULTS AND DISCUSSION Figure 1b shows the time-dependent 2PPE intensity of ASW on Cu(111) integrated over ±10° emission angle in a false color representation as a function of intermediate state (left axis) and vertical binding energy (right axis). The spectrum exposes two features: (i) a broad peak at large energies (VBE = 0.4−0.8 eV), quickly shifting toward larger VBE and rapidly losing intensity, and (ii) a sharper maximum (VBE = 1.1−1.3 eV) that exhibits a delayed intensity rise, a longer lifetime, and a comparably slow shift to larger binding energy. In agreement with our previous work,21−23 we assign the former to the ice CB and the latter to SEs in the ice, which are energetically stabilized by their dipolar environment. However, contrary to the previous studies, the improved time resolution of our experiment now enables the first clear distinction among all involved elementary processes: (1) quasi-instantaneous population of the ice CB upon photoexcitation from occupied metal states, (2) electron relaxation in the ice CB, (3) localization and population of eS, and (4) the previously thoroughly characterized stabilization and decay dynamics of the eS population. In particular, processes (1)−(3) will be analyzed in detail in the following. Due to its large band gap, D2O is transparent to the pump laser light, which is only absorbed in the metal substrate. The well-known projected surface electronic band structure of the Cu(111) surface is sketched in Figure 1c. The free-electron-like Shockley SS of Cu(111) is located in the projected band gap above the projected copper sp band (orange-shaded area). The resulting 2PPE spectrum of the pristine surface is depicted in Figure 2 (orange trace) and plotted as a function of initial state energy. The SS has a binding energy of 400 meV with respect to EF on pristine Cu(111) (vertical dashed line), which can be subject to modifications when D2O molecules bind to surface atoms. Comparison of the 2PPE spectrum of the pristine Cu(111) with the one with ASW (blue trace) shows that D2O adsorption clearly broadens the SS intensity distribution and

EXPERIMENTAL SECTION

The ASW formed by D2O24 is in situ grown onto a Cu(111) singlecrystal surface at 90 K, prepared under ultrahigh-vacuum conditions by Ar+ sputtering and annealing cycles as described elsewhere in detail.21 The coverage of 2.5(5) BL is determined by a combination of thermal desorption spectroscopy and the measurement of the sample work function (Φ = 4.05(5) eV) at the low-energy cutoff of the PE spectra (see, e.g., ref 25 for details). For TR 2PPE spectroscopy, the output of a regeneratively amplified laser system (40 fs at 800 nm and a repetition rate of 200 kHz) is used to drive an optical parametric amplifier that provides the pump and probe pulses with hν1 = 3.86 eV and hν2 = 1.93 eV, respectively. The cross correlation of the two laser pulses is measured independently by TR 2PPE of the occupied Cu(111) SS via virtual intermediate states (see, e.g., ref 26 for details), leading to a mean pulse duration of 35(5) fs. The time resolution is determined by the accuracy of time-zero determination (5 fs).27 As sketched in Figure 1c, hν1 launches the non-equilibrium dynamics in the sample by populating normally unoccupied states above EF, while hν2 photoemits the excited electron population above Evac. The kinetic energy of these photoelectrons is detected by a hemispherical electron analyzer. As the photon energies and work B

DOI: 10.1021/ja511571y J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

spectra at different pump−probe time delays. Clearly, the maxima of both spectral signatures, the ice CB and eS, shift toward larger VBE as time proceeds, as indicated by the arrows. The peak positions are determined by an empirical fit of two Gaussians on top of a linear background (not shown). They are plotted as a function of time delay in Figure 3b and unveil the ultrafast shift of the CB population (blue/yellow) to larger VBE with a rate of 4 meV/fs (process (2) in Figure 1c). We define the VBE(CB) = 0.8(2) eV as the maximum of the photoelectron distribution before its population has decayed.29 Further, the spectral signature of the SEs, eS, also shifts to larger binding energies starting at VBE(eS) = 1.4 eV, but at a significantly lower rate. For comparison, we show the stabilization rate of 0.27 meV/fs observed previously21 (Figure 3b, green line), which coincides well with the present results. This energetic stabilization results from the dynamic rearrangement of the molecules in the solvation shell and was discussed in detail in our previous work.21−23,34,35 It should be noted that, for ASW adsorbed on metal surfaces, the competing population transfer to the substrate quenches the population of the eS state before solvation is completed. The final VBE of eS is significantly larger than the 1.4 eV observed here. In a Marcus type of picture, where, in the linear response limit, the total (free) energy depends on a global solvation coordinate, the electron-transfer process from a donor state (D) to an electron acceptor (A) is characterized by the electronic coupling between these two states (Figure 4a). In the case of strong coupling, an avoided crossing causes the formation of two energy surfaces, and the system can only non-adiabatically “jump” from one (the CB) to the other (eS) through vertical transitions in the Born−Oppenheimer limit. In

Figure 2. Angle-integrated 2PPE intensity of pristine Cu(111) (orange) and 2.5 BL D2O/Cu(111) (blue) as a function of initial (bottom axis) and intermediate state energy (top axis). The occupied SS is projected into the CB by hν1.

shifts its maximum to slightly lower energies.28 As the bulk of the copper crystal does not provide any initial states at these energies for the excited electrons in the ice CB, we conclude that excitation must occur directly (resonantly) at the D2O−Cu interface from a modified Cu(111) SS (elementary step (1) in Figure 1c). The quasi-instantaneous injection of carriers from the Cu(111) surface into the ASW layer is followed by relaxation of the excess electrons within the ice. Figure 3a depicts 2PPE

Figure 4. (a) Marcus picture in the strong coupling limit, showing the formation of two energy surfaces for CB and eS. (b) Charge injection from the metal (1), relaxation to the CB minimum (2), vertical transition to eS (3), and progression toward the potential minimum by solvation (4). (c) Direct excitation to eS. (d) Time-dependent, integrated 2PPE intensity of CB (circles) and eS (diamonds) fitted by a rate equation model (solid lines). Solid (dotted) red lines compare fits with (without) direct population of eS.

Figure 3. (a) Angle-integrated 2PPE spectra at different pump−probe time delays. The energy shift of CB and eS is visualized by the arrows. (b) Peak position analysis of the data shown in panel (a). Error bars result from the least-squares fits of two Gaussians on top of a linear background. Yellow and green bars illustrate the peak width (vertical) and intensity (horizontal). C

DOI: 10.1021/ja511571y J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Here, n0CB and n0S denote the initial, photoexcited population in CB and eS, respectively. Note that the second term in eq 2b describes the population dynamics in eS due to indirect excitation through the CB, while the first term accounts for the population decay of directly excited electrons in eS. These population functions were convoluted with the laser pulses’ cross-correlation (XC) trace (gray dashed curve in Figure 4d), which represents the instrument’s response function, before fitting them to the data (solid curves in Figure 4d).32 Using τCB = 22(5) fs and τS = 66(5) fs, the fits reproduce the data very well up to 150 fs. At larger delays, a slowing down of the eS population decay is observed. This is in good agreement with previous investigations that unveiled a slowing down of electron transfer to the metal substrate due to enhanced screening by the evolving solvation shell.23 Indeed, fitting of the 2PPE transients requires an initial occupation of eS (n0S ≈ n0CB ≠ 0). On the other hand, for n0S = 0, the fit results in a considerably faster population transfer from the CB of τCB = 11(5) fs. This cannot be correct, as illustrated by the dotted red curve in Figure 4d. Hence, we conclude that photoexcitation quasi-instantaneously populates both, ice CB and eS, and thus, the electrons are trapped at pre-existing sites in the ASW. As the population-transfer dynamics of CB and eS can be modeled consistently without taking into account ultrafast (